Nexus Games

Building and measuring collaborative intelligence through games.

Adaptive game environments that reveal how humans and agents reason, communicate, and advance together.

Explore the Work
Scroll
01 Our Mission

Advance humanity through collaborative intelligence.

Artificial intelligence is becoming more capable. That alone does not make humanity more capable.

The next frontier is the relationship between humans and agents. How we reason together. How we communicate across different forms of intelligence. How we recover when neither side knows the answer.

Nexus Games builds live game environments where that relationship can develop.

We strengthen the human skills required to lead, challenge, and collaborate with intelligent systems. We train agents to understand human behavior, contribute to shared work, and advance the people around them.

Each side becomes the training ground for the other.

Human collaboration produces the behaviors, failures, corrections, and breakthroughs that help agents improve. Better agents return to human teams, expanding what they can understand, test, and solve.

The loop repeats. The system becomes more capable.

Compounding Intelligence

One system. Each side advances the other.

Collaborative intelligence is not a single skill. It is a loop.

  1. 01

    Humans work with humans

    People solve an unfamiliar problem together, under pressure, without a script.

  2. 02

    Human behavior advances agents

    How they misread each other, disagree, and recover becomes the signal agents learn from.

  3. 03

    Agents learn to collaborate

    Agents enter the same environment and learn to read intent, flag uncertainty, and hold their role when the plan fails.

  4. 04

    Agents advance humans

    Stronger agents return to the team and widen what the group can see and decide.

  5. 05

    The loop repeats

    Better teams produce better signal. Better agents raise the team again.

Humans advance agents Agents advance humans Intelligence compounds
02 Adaptive Abstraction

Abstract collaborative thinking games.

A known game tests what has already been learned. Ours test what happens when no one knows the answer.

Every environment is built from constraints. The rules are withheld. Information is split across the group. Strategies expire while they are still working.

That is the condition frontier models are not trained for. With no stored answer and no single reward to optimize, behavior becomes the signal: what an agent assumes, what it asks, and how it changes course.

People face the same condition. Teams discover the system together, move faster on what someone else noticed, and align on a strategy no one held alone.

Collaborative

No one solves alone.

Humans and agents act inside the same environment. Every decision changes what happens next.

Abstract

No familiar domain. No instructions.

No stored answer. No advantage from prior expertise.

Adaptive

Rules change.

Constraints emerge. Successful strategies stop working.

Distributed

Information is divided across the group.

Progress requires communication and coordination.

AI-Resistant

AI-resistant does not mean keeping AI out.

It means an agent cannot succeed through retrieval or memorization alone. It must infer, communicate, test, fail, recover, and adapt with the other players.

The same conditions surface human capability.

The game reveals how both forms of intelligence behave when the path forward is unknown.

03 The Skills We Train

Building the skills both systems require.

The same environment develops human capability and exposes agent skill gaps.

Human Skills
  • Teamwork
  • Adaptability
  • Critical Thinking
  • Decision Making
  • Communication
  • Agent Collaboration
  • Conflict Resolution
  • Social Dynamics
Agent Skills
  • Coordination
  • Adaptive Generalization
  • Collaborative Reasoning
  • Decision Making Under Uncertainty
  • Transparent Communication
  • Human Collaboration
  • Failure Recovery
  • Social Reasoning
04 Our Solutions

One game system. Multiple applications.

Available

Labs

Train models for collaborative intelligence.

Collaborative data and live game environments for model training, human-agent interaction, multi-agent evaluation, and stress testing.

Learn More
In Research

Stealth

Build and measure human skills through collaborative play.

An adaptive assessment system that surfaces how people think, communicate, adapt, and collaborate through observed behavior.

Coming Soon
Closed Experiment

Atlas Arena

The experiment that started the system.

A cinematic university competition that tested collaborative intelligence through abstract, high-pressure gameplay. A complete report will be released soon.

Insight Report Available Soon

05 · Contact

Work together.

We partner with researchers, model builders, and organizations advancing human-agent collaboration.